Bank of England governor calls for rigorous AI testing before regulation

Bank of England Governor Andrew Bailey on Wednesday called for rigorous testing of artificial intelligence systems to identify vulnerabilities and establish safeguards before turning to regulation.

In his first article for Substack, Bailey said the risks surrounding AI were “real and increasingly significant,” while stressing that development of the technology should not be halted or prohibited.

“On the contrary, the benefits are immense,” Bailey said, adding that there must be a system for intervention and clear boundaries within which AI operates.

Bailey said important work was already underway in the UK to test AI systems, pointing to the country’s AI Security Institute as part of efforts to develop standards.

“But the pace of progress must accelerate,” he said, adding: “We should proceed with a degree of humility.”

Bailey said testing would inevitably reveal failures and instances in which “models will behave unexpectedly.”

“That is not evidence that testing has failed, rather it is evidence of why testing is necessary,” he said.

Bailey stressed that testing should not be viewed as an alternative to future regulation, saying a more formal regulatory framework could emerge over time.

“But regulation is not, in my view, the right place to start,” he said.

The Bank of England governor specifically highlighted so-called frontier AI – highly advanced, self-learning systems.

Without effective intervention mechanisms, Bailey warned, such systems could become “a closed loop in which the model progressively governs itself.”

US President Donald Trump, meanwhile, has rejected calls for a slowdown, saying the US is leading the AI race ahead of China.

“And, frankly, I want to keep it that way because whoever wins AI, wins,” Trump said.

On Tuesday, after meeting executives from OpenAI, Anthropic, Nvidia, SpaceX, Meta and Google, Trump said they had signed a “morally binding” document that would serve as a “form of protection” against potential AI risks.

Trump also renamed it “Super Intelligence.”

The White House Accord on “Super Intelligence,” described as a “Joint Commitment on Frontier Responsibilities,” calls on companies developing and deploying frontier models to establish internal processes and controls.

The accord says the measures should ensure systems operate as intended and potential problems are identified and addressed quickly.

According to the accord, each participating company should implement four layers of controls and audits covering model development and deployment.